Common Genetic Variants in Proinflammatory and Other Immunoregulatory Genes and Risk for Non-Hodgkin Lymphoma

Profound disruption of immune function is an established risk factor for non-Hodgkin lymphoma. We report here a large-scale evaluation of common genetic variants in immune genes and their role in lymphoma. We genotyped 57 single nucleotide polymorphisms (SNP) from 36 candidate immune genes in 1,172 non-Hodgkin lymphoma cases and 982 population-based controls from a US multicenter study. We calculated odds ratios (OR) and 95% confidence intervals (95% CI) for the association between individual SNP and haplotypes with non-Hodgkin lymphoma overall and five well-defined subtypes. A haplotype comprising SNPs in two proinflammatory cytokines, tumor necrosis factor-α and lymphotoxin-α (rs1800629, rs361525, rs1799724, rs909253, and rs2239704), increased non-Hodgkin lymphoma risk overall (OR, 1.31; 95% CI, 1.06-1.63; P = 0.01) and notably for diffuse large B cell (OR, 1.64; 95% CI, 1.23-2.19; P = 0.0007). A functional nonsynonymous SNP in the innate immune gene Fcγ receptor 2A (FCGR2A; rs1801274) was also associated with non-Hodgkin lymphoma; AG and AA genotypes were associated with a 1.26-fold (95% CI, 1.01-1.56) and 1.41-fold (95% CI, 1.10-1.81) increased risk, respectively (P_trend = 0.006). Among non-Hodgkin lymphoma subtypes, the association with FCGR2A was pronounced for follicular and small lymphocytic lymphomas. In conclusion, common variants in genes influencing proinflammatory and innate immune responses were associated with non-Hodgkin lymphoma risk overall and their effects could vary by subtype. Our results require replication but potentially provide important clues for investigating common genetic variants as susceptibility factors and in disease outcomes, treatment responses, and immunotherapy targets. (Cancer Res 2006; 66(19): 9771-80)

Recent gene expression studies have provided novel insights into the pattern of gene dysregulation in non-Hodgkin lymphoma and established a molecular basis for subclassification of the two major pathologic subtypes, follicular and diffuse large B-cell lymphomas (DLBCL; refs. 1, 2). Although gene expression studies are useful for prognosis, the etiology of non-Hodgkin lymphoma remains enigmatic. Thus far, no high-penetrance gene mutation has been identified for non-Hodgkin lymphoma but a 2-fold increased risk in individuals with first-degree relatives who had non-Hodgkin lymphoma or other hematopoietic disease (3–5) provides strong evidence for genetic susceptibility to non-Hodgkin lymphoma. To date, investigations of immune gene variations in lymphomas have included a limited number of genes, polymorphisms, and non-Hodgkin lymphoma outcomes (6–11). Recent gene expression studies further implicate host inflammatory responses (12) and the nuclear factor-κB (NF-κB) pathway, thus supporting the conceivable role for polymorphic alleles of key immunoregulatory genes and the risk for non-Hodgkin lymphoma (13).

The immune response is represented by a large number of overlapping pathways that include cytokines, chemokines, cell adhesion molecules, interferons, and innate immune response molecules. The coordination of signals required to preserve the balance within this robust and pleiotropic network must be maintained in healthy individuals. Sustained perturbation of this balance, such as that caused by inherited gene mutations, can result in significant disease [e.g., Janus-activated kinase (JAK) 3 or interleukin (IL)-7 receptor A deficiency and severe combined immunodeficient disease (SCID); ref. 14]. Common genetic variants can also alter the expression or function of key genes, disrupting the balance cytokines and cells on which various cytokines act [T-helper (Th) 1, Th2, and T-regulatory cells; ref. 15]; these common variants have been associated with outcomes in autoimmune disorders (e.g., lupus), cancer (e.g., gastric), and infectious diseases (e.g., Helicobacter pylori; refs. 16–18).

Given the substantial weight of evidence for immune function in non-Hodgkin lymphoma, we evaluated 57 single nucleotide polymorphisms (SNP) in 36 proinflammatory and other immunoregulatory genes and non-Hodgkin lymphoma risk. Variants were selected based on a priori laboratory evidence suggesting functional consequences for an allele, associations in previous studies in autoimmune disorders, cancer or infectious diseases, or for additional gene coverage in haplotype analysis (Table 1). We present results for non-Hodgkin lymphoma overall and five subtypes: DLBCL, follicular lymphoma, small lymphocytic lymphoma (SLL), marginal zone lymphoma, and T-cell lymphoma.

The study population has been described previously in detail (4). We included 1,321 newly diagnosed non-Hodgkin lymphoma cases identified in four Surveillance, Epidemiology, and End Results (SEER) registries (Iowa; Detroit, MI; Los Angeles, CA; and Seattle, WA) ages 20 to 74 years between July 1998 to June 2000 without evidence of HIV infection. Population controls (1,057) were identified by random digit dialing (<65 years) and from Medicare eligibility files (≥65 years). Overall participation rates were 76% in cases and 52% in controls; overall response rates were 59% and 44%, respectively. Written informed consent was obtained from each participant before interview. All study participants were asked to provide a venous blood or mouthwash buccal cell sample. We obtained blood samples from 773 cases and 668 controls and buccal cells from 399 cases and 314 controls. We evaluated the 1,172 (89%) cases and 982 (93%) controls for whom biological samples were obtained for genotyping (Table 2). Genotype frequencies for individuals who provided blood compared with buccal cells were equivalent (19).

Each registry provided non-Hodgkin lymphoma pathology and subtype information derived from abstracted reports by the local diagnosing pathologist. Although final translation to International Classification of Diseases for Oncology (ICD)-O-3/WHO classification is on-going, all cases were histologically confirmed and have been coded according to the ICD, 2nd edition (20), which incorporates both the revised European-American classification of lymphoid neoplasms and working formulation and translates directly to the ICD-O-3/WHO. We evaluated the following histologic outcomes: (a) non-Hodgkin lymphoma overall, (b) B-cell lymphomas, (c) T-cell lymphomas, (d) DLBCL (B-cell subtype), (e) follicular lymphoma (B-cell subtype), (f) marginal zone lymphoma (B-cell subtype), and (g) SLL (B-cell subtype). Of note, 28 SLL cases were later identified by pathology review as chronic lymphocytic leukemia (CLL). Because SLL and CLL comprise the same disease and differ only in initial presentation, we refer to these as SLL for the purposes of this article.

DNA extraction. DNA was extracted from blood clots or buffy coats (BBI Biotech, Gaithersburg, MD) using Puregene Autopure DNA extraction kits (Gentra Systems, Minneapolis, MN). DNA was extracted from buccal cell samples by phenol-chloroform extraction methods (21).

Genotyping. We selected SNPs with ≥5% prevalence with evidence of functional consequence or association with non-Hodgkin lymphoma in human studies. Most of our 57 SNPs in 36 immune genes met these criteria (Table 1); SNPs that did not meet the criteria were included for haplotype analysis (22). Eleven polymorphisms were included in on-going consortial efforts as shown in Table 1; we note that consortial efforts were limited to Whites and two subtypes (DLBCL and follicular lymphoma). Here, we present additional results by cell lineage (B-cell versus T-cell lymphomas) and, for two additional B-cell subtypes (SLL and marginal zone), show haplotype analyses that include additional SNPs in key loci and include genotype results for Black study subjects, all of which are presented for the first time. Further, all results from the remaining 46 SNPs are shown here for the first time.

Genotyping was conducted at the National Cancer Institute Core Genotyping Facility (Gaithersburg, MD) using the Taqman (Foster City, CA) or Epoch (Bothell, WA) platforms. Sequence data and assay conditions are provided online (23).⁸

Genotyping results in blood-based DNA samples were analyzed first. For SNPs with significant (P < 0.05) or suggestive associations (borderline significance or trend observed), genotyping was completed in buccal cell samples, which in general had less DNA than blood-based samples. SNPs informative for constructing haplotypes were also genotyped in buccal cell samples. We had complete data from all participants for 42 SNPs.

Quality control. Replicate samples (40) from two blood donors each and duplicate samples from 100 participants processed in an identical fashion were interspersed for all assays and blinded from the laboratory. Agreement for quality control (QC) replicates and duplicates was ≥99% for all assays. For each plate of 368 samples, genotype-specific QC samples were also included and comprised four each: homozygote wild-type (WT), heterozygote, homozygote variant, and DNA-negative controls.

Successful genotyping was achieved for 96% to 100% of DNA samples for all SNPs; completion rates did not differ by blood- or buccal-based DNA. One SNP (CXCL12, rs1801157) in black controls was not in Hardy-Weinberg equilibrium (P < 0.01); genotype assignments and QC data from replicates and duplicates were thus rechecked for this SNP and the accuracy of this assay was confirmed, per sequence and assay specifications on the SNP500 Web site.⁸

Gene-disease associations. We calculated odds ratios (OR) and 95% confidence intervals (95% CI) for each genotype with each non-Hodgkin lymphoma outcome, using the homozygous WT genotype as the referent group. We conducted stratified analysis by age (<60 and ≥60 years), sex (male and female), and race (Whites and Blacks). Finding no significant differences in the risk estimates by each of these strata, we pooled the results and adjusted for the study design variables: age (<54, 55-64, and 65+ years), sex, and race (White, Black, and other). For each outcome, we calculated the P_trend based on the three-level ordinal variable (0, 1, and 2) of homozygote WT, heterozygote, and homozygote variant in a logistic regression model. All logistic regression models were unconditional and conducted using SAS version 8.2 (SAS Institute, Cary, NC).

Because some of our results could be false-positive findings due to chance, we therefore calculated the probability that our findings were false positives using two methods. We used the Benjamini-Hochberg method to calculate the false discovery rate (FDR; ref. 24), which reflects the expected ratio of false-positive findings to the total number of significant findings. We applied the FDR method to the Ps for trend as this allows for the fewest number of comparisons and thus degrees of freedom to assess the additive model. Briefly, the additive model (or trend test) measures the risk with each additional variant allele. For evaluating the FDR for B- and T-cell lymphomas, both sets of P_trend were included in the calculation; for evaluating subtypes for FDR, all subtypes were evaluated simultaneously to account for all subtype comparisons made (e.g., DLBCL, follicular, SLL, and marginal zone). Because the FDR does not consider prior probability, we also calculated the false-positive report probabilities (FPRP; ref. 25). FPRP takes into account our prior probabilities, which we specify as a range (0.001-0.1) and applied to all findings. We used a criterion of 0.20, as suggested by the original publication on the method. Using this criterion, we would therefore interpret a FPRP of 0.20 of having a 20% probability of being a false-positive result.

Haplotype analysis. Within non-Hispanic White controls, haplotype structures were examined using Haploview version 3.11 (26). Two well-characterized haplotypes were investigated: (a) IL-10 (rs1800871, rs1800872, and rs1800896; ref. 27) and (b) TNF/LTA (rs909253, rs2239704, rs1800629, rs361525, and rs1799724; ref. 28). Although haplotype blocks were constructed for IL-1, IL-4, and IL-6, their high correlation (r² ≥ 0.91) in our population provided low frequencies of haplotypic variants for further analysis. We estimated haplotypes using the expectation-maximization algorithm (29). Using the statistical package, HaploStats in software R (version 2.0.1; ref. 30), overall differences in haplotype distribution between each non-Hodgkin lymphoma subtype and controls were assessed using the global score test (31). Risk estimates were estimated from the additive model, which fitted a logistic regression model and used posterior probabilities of the haplotypes as weights to estimate the regression coefficients in an iterative manner (31), adjusting for age and sex.

We present all results of our investigation of common genetic variants for non-Hodgkin lymphoma overall and for each non-Hodgkin lymphoma subtype (Supplementary Tables S1-S4). We also show results restricted to Whites (Supplementary Tables S5-S8).

Proinflammatory response genes. In our analyses of all study participants (Whites and Blacks), polymorphisms for the proinflammatory cytokines, tumor necrosis factor (TNF; G-308A, rs1800629), and lymphotoxin-α (LTA; A252G, rs909253) were associated with increased risks for non-Hodgkin lymphoma and risks were particularly elevated for DLBCL, as recently reported among Whites. We further found statistically significantly 4- and 3-fold risk increases for T-cell and marginal zone lymphomas, respectively (Fig. 1; Supplementary Tables S2 and S4). Among Blacks, consistently increased risks for non-Hodgkin lymphoma overall and DLBCL were observed (TNF G-308A and non-Hodgkin lymphoma: OR_GA, 1.5; OR_AA, 5.7; P_trend = 0.08). Because single polymorphisms may not fully reflect gene function, results of our haplotype analysis are particularly noteworthy. Specifically, we found the haplotype that included both TNF G-308A and LTA A252G risk alleles, LTA C-91A, LTA A252G, TNF C-857T, TNF G-308A, and TNF G-238A (C-G-C-A-G), conferred increased risk for non-Hodgkin lymphoma overall (OR, 1.31; 95% CI, 1.06-1.63), DLBCL (OR, 1.64; 95% CI, 1.23-2.19), and T-cell lymphomas (OR, 1.69; 95% CI, 1.00-2.86) when compared with the common A-A-C-G-G haplotype (Table 3). Risk estimates were not statistically significantly elevated for marginal zone lymphomas (OR, 1.44; 95%, 0.86-2.41).

The IL-8 receptor B (IL-8RB; rs1126580) 3′-untranslated region (UTR), Ex3−1010G>A SNP was associated with a decreased risk for non-Hodgkin lymphoma (OR_AG, 0.77; 95% CI, 0.61-0.97; OR_GG, 0.76; 95% CI, 0.55-1.05); the trend for DLBCL (OR_AG, 0.82; OR_GG, 0.45; P_trend, 0.007) was statistically significant.

Other immunoregulatory genes. Of Th1/Th2 cytokine polymorphisms evaluated, statistically significant associations were limited to specific subtypes. The IL-4 receptor (IL-4R; C-29429T, rs2107356) homozygous variant polymorphism increased risk for DLBCL (OR_TT, 1.66; 95% CI, 1.13-2.42; Supplementary Table S3). Conversely, IL-15 receptor α (IL-15RA, rs2296135) was associated with a decreased risk for follicular lymphoma (OR_GT, 0.63; OR_TT, 0.54; P_trend = 0.008). Statistically significant increased risk for SLL was observed for IL-12B (IL-12B, rs3212227: OR_AC, 1.40; OR_CC, 1.99; P_trend = 0.02) and IL-13 (IL-13 Q144R, rs20541: OR_AG, 1.45; OR_AA, 2.17; P_trend = 0.01 and IL-13 C1069T, rs1800925: OR_CT, 1.41; OR_TT, 1.92; P_trend, 0.02; Supplementary Table S4). For marginal zone lymphomas, the JAK3 (rs3008) polymorphism significantly increased risk (OR_CT, 2.43; 95% CI, 1.13-5.25; OR_TT, 3.59; 95% CI, 1.61-8.02; P_trend = 0.002; Supplementary Table S4).

For non-Hodgkin lymphoma overall, a significant increased risk and trend were observed for the low-affinity Fcγ receptor 2A (FCGR2A; H165R, rs1801274), which couples cellular and humoral immunity (OR_AG, 1.26; 95% CI, 1.01-1.56; OR_AA, 1.41; 95% CI, 1.10-1.81; P_trend = 0.006; Fig. 2; Supplementary Table S1). Associations were consistent for both B- and T-cell lymphomas (B cell: OR_AG, 1.28; OR_AA = 1.40; P_trend = 0.01; T cell: OR_AG, 1.75; OR_AA, 2.31; P_trend = 0.03; Supplementary Table S2). Among B-cell lymphoma subtypes evaluated, increased risks were observed for all and statistically significant for follicular lymphoma and SLL (follicular: OR_AG, 1.48; 95% CI, 1.03-2.12; OR_AA, 1.52; 95% CI, 1.01-2.26; SLL: OR_AG, 1.55; 95% CI, 0.97-2.48; OR_AA, 1.65; 95% CI, 0.98-2.77; Supplementary Tables S3 and S4). We observed no associations between the chemokines CCR2 and CCR5 and our non-HIV-associated non-Hodgkin lymphoma cases. We further observed no significant gene-gene interactions with any of the aforementioned SNPs above what was expected in an additive model.

In aggregate, our results suggest that perturbations in inflammation stemming from common genetic variants in proinflammatory cytokines TNF and LTA could contribute to the development of non-Hodgkin lymphoma. Our results extend previous associations from smaller studies and those with prognostic outcomes that have suggested the importance of TNF in lymphoma and other inflammatory conditions, including rheumatoid arthritis and Sjogren's syndrome (9, 11, 32–34). We extend these associations between the individual TNF G-308A and LTA A252G polymorphisms to DLBCL, marginal zone lymphoma, and T-cell lymphoma subtypes. Our association for TNF G-308A and DLBCL among Blacks extends recent consortium results found among Whites (22). Analysis in the InterLymph case-control consortium, for which we contributed data from our White population for these two polymorphisms for DLBCL and follicular lymphomas, strongly supports the single SNP association between the TNF G-308A polymorphism and increased risk for DLBCL (22). Replication of our findings for Blacks as well as for marginal zone lymphoma and T-cell lymphoma subtypes is required.

We further report our analysis of five polymorphisms within the TNF/LTA locus for a more detailed haplotype analysis. Because the analysis of single common polymorphisms does not necessarily capture the extent of genetic variation across a locus or gene function (32, 35), our haplotype findings are instrumental in showing that this locus could be associated with non-Hodgkin lymphoma and that an untested SNP in linkage disequilibrium could be the causal variant(s) (28). We show the importance of this locus in all non-Hodgkin lymphoma and specifically for DLBCL and T-cell lymphomas. Both TNF G-308A and LTA A252G have been shown to alter expression of TNF-α (35) and LTα (36), respectively. Our observations are particularly notable because promoter variants of TNF, in concert with the LTA haplotype, have been associated with increased gene expression and shown to play pivotal roles in inflammation. TNF-α, the protein product of TNF, activates the NF-κB pathway, a hallmark of inflammation (37, 38). LTA, a member of the TNF family and colloquially known as TNF-β, is also a critical signal in the inflammation cascade, contributing to antiviral activities. Because TNF functions as an activator of NF-κB, we believe further investigation of the role NF-κB plays in lymphoma may provide mechanistic clues for etiology, disease outcome, and treatment and contribute to a more complete understanding of lymphomagenesis (39). Our results further support the findings from complementary gene expression studies, which have implicated host inflammatory responses (12) and the NF-κB pathway (13) in DLBCL.

We observed a common variant in the chemokine IL-8Rβ (16% homozygotes) associated with decreased risk of non-Hodgkin lymphoma and specifically DLBCL and T-cell lymphomas. IL-8RB is the receptor for the most potent chemokine, IL-8, which induces chemotaxis of neutrophils; although variants in IL8 have been associated previously with gastric cancer and infectious pathogens (40, 41), our data implicate a role for its receptor in lymphomagenesis. We also observed an increased risk between the low-affinity FCGR2A nonsynonymous SNP, H165R, and non-Hodgkin lymphoma. This SNP is particularly interesting because of its effect on-binding of IgG on the surface of phagocytes, a critical step in innate immunity (42–44). The codominant histidine and arginine alleles are functionally relevant as high and low binding, respectively, leading to differential handling of IgG complexes (45). This example of a key gene in the innate immune response increased risk for follicular lymphoma and SLL.

To pursue our findings, we believe that a dense haplotype analysis that saturates specific immune genes is warranted. Because TNF and LTA are located within the major histocompatibility region, detailed haplotype structure spanning chromosome 6 in both association studies and in vitro correlative studies designed to identify the causal SNP(s) should proceed, as should investigation of the human leukocyte antigen (HLA) and killer immunoglobulin receptor family, for which HLA is a ligand (46, 47). Haplotypes are particularly useful for assessing genetic variability in larger genetic regions, such as HLA, and further diplotype analyses should be considered so that data on the coexisting second allele are evaluated. Furthermore, it remains plausible that there may exist additional genetic associations in the presence of a relevant exposure (i.e., gene-environment interaction; ref. 48) and the effect of additional post-translational modifications and other functional alterations in modulating immune responses would not be captured in our results (49).

Other findings of note include the association between IL-4R and increased DLBCL risk. IL-4R significantly up-regulates receptor responses to IL-4, resulting in increased proliferation, IgE production, and atopic phenotypes (50). For DLBCL, our results suggest further evaluation of a Th2 response, of which IL-4 is a signature cytokine. Interestingly, we observed a decreased risk for follicular lymphoma with alteration in IL-15RA, the high-affinity binding receptor for IL-15. IL-15 and IL-15RA on follicular dendritic cells are a key component to germinal center B-cell survival and proliferation (51). Our results suggest that alterations in IL-15RA may thus adversely affect B-cell survival and proliferation and decrease follicular lymphoma risk. Increased risks for SLL were found with IL-12B and IL-13 polymorphisms. Although IL-12B engages cell-mediated immunity against infections and pathogens, IL-13 engages the humoral immune response (52). Because both down-regulate proinflammatory cytokines, these results might suggest a less important role for inflammatory cytokines and SLL. Finally, increased risk with marginal zone was observed with JAK3. Although JAK3 has been implicated in anaplastic large cell lymphoma and mantle cell lymphoma, its role and apparent specificity to marginal zone is unclear.

Study strengths include the population-based design, large sample size and adequate power to detect main effects of common genes for non-Hodgkin lymphoma risk and reduce the potential for false-positive and false-negative findings (53). The associations we observed for the TNF/LTA haplotype and for FCGR2A were noteworthy by the alternative false probability report probabilities methods (25) with a 0.1 to 0.001 range of prior probabilities. Limitations of our study include loss of eligible subjects to death, illness, and refusal to participate; it is thus possible that our case group would have been biased away from the more aggressive forms of disease. A recent study that included data from our study found no differences in genotype frequency by participation status beyond that expected from chance alone (19). Potential limitations of any study investigating multiple SNPs include false-positive findings. We have thus evaluated our results in consideration of their probability of being a false positive by calculating the FDR (54) and FPRP (25) as described in the Methods. The FDR values for DLBCL based on the P_trend of the TNF G-308A and LTA A252G variants were both 0.05, taking into account all SNPS tested for association with risk of each subtype evaluated in this report (DLBCL, follicular, SLL, and marginal zone); the FPRP value was 0.1 for an additive model with an OR of 1.3 and thus below our criterion of 0.2. In other words, by FPRP, the additive model for the TNF G-308A and LTA A252G variants have a 10% probability of being a false positive. Therefore, by both FDR and FPRP, these associations have only a small probability of being a false positive. The FDR value of the FCGR2A H165R variant was 0.10 for all non-Hodgkin lymphoma and the FPRP value was 0.16, also indicating that there is a low probability of this association being a false positive. Recent concerns for population stratification (55) also prompted analyses stratified by race and by study site and we found our results consistent by race and by study site. For non-Hodgkin lymphoma subtypes, there may have been inadequate power to detect modest associations for rare genotypes and less common subtypes; confirmation of all subtype-specific results in additional studies will therefore be necessary. Finally, although a large study, our evaluation comprises a small proportion of these genes as permitted by our candidate SNP selection process, which was based largely on available biological evidence and validated assays.

In conclusion, our data support a role for common immune gene variants in modulating risk for non-Hodgkin lymphoma. Specifically, our data indicate a clear role for proinflammatory genes in DLBCL and T-cell lymphomas as well as other immunoregulatory genes, such as innate immune genes in non-Hodgkin lymphoma. If confirmed in other large studies and in pooled analyses, our data suggest distinct lines of inquiry for investigating the etiology of non-Hodgkin lymphoma and its subtypes. The identification of genes where genetic variations could alter expression levels focuses attention on genes that could be suitable targets for preventive or therapeutic strategies. It is plausible that propensity for inflammation resulting from common genetic variants would likely be chronic and further contribute to DNA damage. Our results complement recent advances in microarray analyses in the molecular profiling of DLBCL (2, 12, 56) and follicular lymphoma (1, 57). Because those data represent alterations in expression level at a single time point, we suggest that a series of imbalances in the network of genes that control inflammation could have deleterious consequences and contribute directly to lymphomagenesis (13, 58).

Grant support: Public Health Service contracts N01-PC-65064, N01-PC-67008, N01-PC-67009, N01-PC-67010, and N02-PC-71105.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank our SEER collaborators for the recruitment and conduct of the study's field effort and collection of biological specimens; Robert Welch and Sunita Yadavalli (NCI Core Genotyping Facility) for their tremendous efforts in the specimen handling and laboratory analysis of genotyping data; and Peter Hui (Information Management Services, Inc., Silver Spring, MD) for programming support.